I quite like these smaller niche events because they allow you to meet network planners and other analysts, and discuss issues on a more personal basis. MNO is on the smaller side of these with about 40-50 attendees, mostly from Europe but a few from Africa and Asia. As with almost every conference, the introduction starts off talking about 1000x data traffic growth. Both the presentations and side conversations drilled down into what's really going on in networks, which varies considerably in different parts of the world.
Characterising the data growth – where, when and how?
I asked a couple of presenters if they could be more specific about where this 1000x growth is coming from. Will it be in the urban areas or rural/remote – surely both would grow proportionately in density? Is it more related to in-building or stationary use cases rather than from fast moving cars and trains?
I'm not convinced that anyone knows the future particularly better than you or I do. I'm often answered with a comment such as "I left my crystal ball at home today". Most focus on the urban areas – some on the streets outside, others in-building. One insight from Andy Sutton of EE, who pointed out that those in rural areas are generating the same kind of traffic profile on 4G that you see on residential broadband – once connected to the internet they want to make use of it. So while the most dense traffic might be seen in urban areas because of the population density, the growth rates could equally apply elsewhere. Farmers have the same needs as everyone else, and perhaps they like to do internet banking & shopping just as much as the rest of us.
High Definition Video will be the major component of data growth
Andy was also quite specific with his view that it will be high definition video that will drive the need for urban small cells in the long term. Where a 2Mbps standard rate video stream is common today, HD video can use up to 18 Mbps. With data charged by the Gigabyte, it will be up to consumers to decide if they want to pay the extra 10x cost. EE only have one legacy unlimited data tariff left, most of their customers are on bundled tariffs, so they (EE) wouldn't lose out.
The use of mobile for video may seem strange, but Germany is already considering turning off DVT-B and would rely only on cable and satellite. Both China and India have crossed the rubicon where more data is sent by mobile than fixed networks. Elsewhere in Europe, analogue TV is still in use which causes great problems for the Croatian operator VIPnet who is trying to roll out a 4G network using the same spectrum.
Peter Zidar of the UMTS Forum points out that the new 4G data services can have lower latency and faster speeds than many of today's hard disk drives, meaning that we might achieve a faster and more responsive service when online rather than offline. He concluded with four predictions for the future:
- Small cells will be key to improve capacity
- Video consumption will shift away from broadcast
- The industry will need more spectrum
- Remote data may not seem so remote anymore
Forget bits per Hertz per second: It's more about efficient applications now
Andy Sutton was also vocal about how optimisation of Moble Networks has moved on from focussing on the pure air interface bit rate. Operators measure the individual end user experience of specific popular apps such as Facebook, YouTube etc.
Optimising a mobile network today involves complete end-to-end analysis to avoid bottlenecks and reduce latency. It also needs top to bottom analysis of the protocol stacks to see where the overheads lie. This is where the discrepancy between Application bits/second and radio bits/second are most evident. It seems a waste of all that expensive spectrum to add in many layers of packet headers only to unwrap them at the other end.
So it's not Bits per Hertz per second anymore, it's application bits per second per Hertz. Fundamentally, we're using a 21st Century 4G network to deliver IP protocols designed in the 1960's for use over dodgy and unreliable networks. While IPv6 might have solved the problem of limited address space, it didn't really deliver much more. We need better protocols at the application levels that make best use of the underlying reliable low latency network.
EE's network is quickly becoming a world leader. We'll soon have 300Mbps service across London, delivered by combining 20MHz at 1800MHz and 20MHz at 2600. A further 15MHz of 2600 spectrum will increase speeds up to 410Mbps and is really only held back by the lack of suitable devices. It's this huge stockpile of spectrum that allows EE to postpone widespread deployment of Small Cells but these are still seen as inevitable when the time is right. London could move ahead of Korea, US and Japan with peak LTE speeds faster than anywhere else.
Signalling is less of a problem but still significant
Michael Fritz, an independent consultant, gave a detailed view of current traffic profiles. It seems that while tablets and laptops are popular, they're most often used with Wi-Fi leaving Smartphones as the big consumer of mobile bandwidth. Many users like LTE so much they switch off Wi-Fi completely.
Signalling data, which isn't charged for, continues to grow. A measure of Diameter signalling shows that it doubled annually, driven by social media and similar apps that perform frequent status checks but transmit little or no data. The signalling overloads generated by the Fast Dormancy feature(where smartphones can go to sleep while maintaining the data connection) was effectively solved by URA in Release 8 which is now incorporated in most smartphones. This reduced the number of signalling messages from 30 to 7 each time a device wakes up.
Urban densification will be driven on a sector by sector basis.
Stefan Teral re-iterated a common industry view that each macrocell will be augmented by no more than 3-5 small cells. Andy Sutton clarified the situation with a couple of significant observations:
- 3G wasn't designed for HetNet or Co-Channel mode, sharing the same frequency as the macrocell. 4G specifically included this from the outset, co-ordinating the small cells and macrocell with features like ABS and eICIC to ensure they work smoothly together.
- Adding a single 4G urban small cell alongside a macrocell probably doesn't make sense because it wouldn't add much capacity. If you were going to deploy them, you'd want to use several to add significant capacity. This would be done on a sector by sector basis – some sectors might have 3 to 5 (or more) while others (e.g. serving a fast moving highway) might have none.
3G metrocells still have a place in some parts of the world
I spoke with a couple of engineers from Sri Lanka, where the national operator provides 2G, 3G and 4G service. Most of their data customers are on 3G and will remain so. To address this, they are deploying urban 3G metrocells using equipment from Huawei and ZTE. They were well aware of the issues of 3G co-channel interference with the macrocell layer and have been able to allocate dedicated spectrum for the small cell layer.
So while I hear a common story in developed countries that urban small cells will be 4G only (for the reasons Andy stated above), it's clear that there are countries where 3G traffic demand would be better addressed by 3G urban small cells. The story is different for in-building, where high quality 3G voice service remains in the strong demand almost everywhere.
A few other snippets
Exhibitors included Systemics, who told me that operators benefit from a thorough 3rd party audit of their radio equipment and utilisation. Often they find a lot of relatively underused kit in parts of the network which can be removed and reused elsewhere, saving significant amounts of money. I also spoke with Equiendo, who conduct drive and walk testing. Apparently you can tell quite a lot about what your competitors are doing (frequencies allocated, features used, quality/speeds delivered) just by walking around. Neither had seen much activity yet for small cell deployments and had not developed any specific small cell oriented features or services.
You may have heard that the new iPhone 6 includes Wi-Fi calling as a built in feature. This will be used by EE, who had announced they were trialling this to replace their Wi-Fi UMA service. I believe it's intended for use in remote locations when out of coverage rather than as a money saver. Other operators, such as Three, require you to use a dedicated App, making the experience quite clunky and awkward to use. I was unable to ascertain when any UK operators will launch VoLTE although it's clear trials are ongoing.